Do Shocks Have a Significant Impact on Elections?

This week, I will be doing the first blog extension. I will be replicating the Dobbs NYT example from section, but with the shock of recent inflation news. I predict that rising coverage on the state of the economy and inflation would have a negative affect on the President’s party (Democrats) for generic ballot support. I am unsure whether or not this would also translate to an increase in support for Republicans but I am certain that I predict to see some hesitation towards Democrats due to economic inflation and market performances. I will be using the NYT developer API to source my data for news coverage.

I will also try to modify and revise my existing predictive model from week 6.

Background Literature on Shocks and Elections

There is varying literature on the effects of shocks on elections. Achen and Bartels (2017) and Healy and Malhotra (2010) found that natural disasters such as shark attacks and tornados decreased support for incumbents. Further research in elections has also found some correlation to college football team losing their game and a subsequent decreased support for incumbents (Healy, Mo, Malhotra 2010). However, it is important to note that most of these relationships are found to be relatively weak. For this reason, I am skeptical to heavily rely on shocks and scandals to inform my model.

New York Times Coverage on Biden and Inflation

Below is a plot of how the coverage of the terms ‘inflation’ + ‘Biden’ have changed from the period between January 2022 until October 15th, 2022. We can clearly see there is a huge spike in the months of August between weeks 31-33. I am looking at the first 15 pages of coverage from the NYT API. This equates to 160 articles out of a total of around 1000 written.

In the next plot, I will compare how this coverage lines up with general trends in the general ballot support for Democrats and Republicans in the same time period.

Peak Inflation Coverage Lines Up With Unexpected Drops in Support

Above is a graph that compares the peak weeks of coverage as evidenced in the plot above (Weeks 31-33) to the results of general ballot support pollings in the same time period.

Inflation and Biden News Punishing the Republicans More?

An interesting observation of this graph in particular is how the dates line up perfectly with an overall drop in approval ratings for both Democrats and Republicans. However, the drop is much more significant for the non-incumbent party (Republicans) rather than President Biden’s party. One possible explanation for the drop in both Democratic and Republican support could be that the coverage on inflation critiqued both sides, causing overall disillusion with the state of government and the economy. Interestingly enough, Republicans seem to be more punished than Democrats in this data. This may be explained or investigated further by creating sentiment scores for how the coverage was portraying Biden’s response as either positive, negative, or neutral. Furthermore, the last thing to consider is that correlation does not equate to causation so there may be a possibility than other (possibly more important or popular) news coverage that I didn’t consider in these weeks impacted Americans’ support of the parties.

My Existing Model and Predictions

My current model uses several inputs including local level economic data such as state unemployment, inflationary levels that are standard across the states, incumbency status for each district, ad spend by candidate, and expert polling averages.

Why I Am Not Including Shocks Into My Model…

This week, I am choosing to not include data on shocks. My reasoning for this is that firstly, I am unsure of how exactly to quantify shock value as a variable of input into my model. Secondly, I believe that generic ballot polling averages do just as good of a job as accounting for shocks. The result of change in popular support is what we are most interested in and that is already accounted for.

Therefore, this week I am choosing to keep my model intact with some minor adjustments to bettter improve predictive power such as including a confidence interval.

Below are my confidence intervals for my prediction
Observations 7320
Dependent variable DemVotesMajorPercent
Type Linear regression
𝛘²(5) 1751181.50
Pseudo-R² (Cragg-Uhler) 0.74
Pseudo-R² (McFadden) 0.16
AIC 53229.66
BIC 53277.95
Est. S.E. t val. p
(Intercept) 90.89 1.13 80.18 0.00
avg -6.50 0.05 -139.60 0.00
Unemployed_prct 0.41 0.20 2.05 0.04
winner_candidate_incIncumbent 3.52 0.25 13.95 0.00
Receipts -0.00 0.00 -8.95 0.00
turnout -26.95 1.37 -19.66 0.00
Standard errors: MLE
##                                          2.5 %            97.5 %
## (Intercept)                    88.663879077489  93.1073222121294
## avg                            -6.587490243470  -6.4050808572814
## Unemployed_prct                 0.017351853208   0.7937076706067
## winner_candidate_incIncumbent   3.025715186269   4.0148971522350
## Receipts                       -0.000001478649  -0.0000009472493
## turnout                       -29.633723225312 -24.2598668661472

Final Prediction for Week 7

My final prediction for Week 7 is below

Final Thoughts and Predictions

My final prediction for week 7 is the following:

References

Christopher H Achen and Larry M Bartels. Democracy for Realists: Why Elections Do Not Produce Responsive Government, volume 4. Princeton University Press, 2017. URL https://hollis.harvard.edu/primo-explore/fulldisplay?docid=TN_ cdi_askewsholts_vlebooks_9781400888740&context=PC&vid=HVD2&search_scope= DRAFT: everything&tab=everything&lang=en_US.

Marco Mendoza Avin ̃a and Semra Sevi. Did exposure to COVID-19 affect vote choice in the 2020 presidential election? Research & Politics, 8(3): 205316802110415, July 2021. ISSN 2053-1680, 2053-1680. doi: 10.1177/ 20531680211041505. URL https://hollis.harvard.edu/permalink/f/1mdq5o5/TN_cdi_doaj_primary_oai_doaj_org_article_f43f65041eb14d4f839740deb9063b43.

Andrew Healy, Neil Malhotra, et al. Random events, economic losses, and retrospective voting: Implications for democratic competence. Quarterly Journal of Political Science, 5 (2):193–208, 2010. URL https://hollis.harvard.edu/primo-explore/fulldisplaydocid=TN_cdi_crossref_primary_10_1561_100_00009057&context=PC&vid=HVD2&search_scope=everything&tab=everything&lang=en_US

Anthony Fowler and Andrew B Hall. Do Shark Attacks Influence Presidential Elections? Reassessing a Prominent Finding on Voter Competence. The Journal of politics, 80(4): 1423–1437, 2018. ISSN 1468-2508. URL https://hollis.harvard.edu/primo-explore/9fulldisplaydocid=TN_cdi_crossref_primary_10_1086_699244&context=PC&vid=HVD2&search_scope=everything&tab=everything&lang=en_US